Childhood overweight and obesity has increased at an alarming rate and reached epidemic proportions throughout the world over the last decades, although great regional differences have been detected( Reference Moschonis, Tanagra and Vandorou 1 – Reference Haug, Rasmussen and Samdal 6 ). Different earlier studies, as well as the first published reports from the ENERGY (‘EuropeaN Energy balance Research to prevent excessive weight Gain among Youth’) project, indicate that childhood overweight and obesity is significantly and substantially more prevalent among children from lower educated and ethnic minority parents, as well as in southern compared with northern European countries( Reference Brug, van Stralen and Chinapaw 7 , Reference Brug, van Stralen and Te Velde 8 ).
The increased prevalence of overweight and obesity among children and adolescents can mainly be attributed to certain behaviours related to diet and physical activity( Reference Shrewsbury and Wardle 9 , Reference Silventoinen, Rokholm and Kaprio 10 ). Literature on the association between eating patterns and overweight indicates that breakfast habits may be of importance( Reference Szajewska and Ruszczynski 11 ). The results of several cross-sectional studies across the world suggest that skipping breakfast is related to higher BMI and increased probability of being overweight/obese( Reference Haug, Rasmussen and Samdal 6 , Reference Tzotzas, Kapantais and Tziomalos 12 – Reference Affenito 15 ). This finding is also supported by the ENERGY study data( Reference Bjørnarå, Vik and Brug 16 ) and by prospective studies( Reference Timlin, Pereira and Story 17 , Reference Niemeier, Raynor and Lloyd-Richardson 18 ). Despite the probable importance of breakfast consumption, breakfast skipping is common among children and adolescents( Reference Tzotzas, Kapantais and Tziomalos 12 , Reference Rampersaud, Pereira and Girard 14 , Reference Horikawa, Kodama and Yachi 19 , Reference Hallstrom, Vereecken and Ruiz 20 ). Furthermore, breakfast habits adopted in childhood may track into adulthood( Reference Merten, Williams and Shriver 21 , Reference te Velde, Twisk and Brug 22 ). Hence, the identification of those factors related to overweight/obesity in children and of the paths via which these associations occur, e.g. via breakfast habits, is important in order to design more effective, tailor-made interventions.
Children’s eating habits are associated with socio-cultural, economic and physical environmental factors. Previous studies examining the association between socio-economic status (SES) and breakfast habits indicate that a child or adolescent living in a single-parent family( Reference Vereecken, Dupuy and Rasmussen 23 ) or in a family with low SES( Reference Timlin, Pereira and Story 17 , Reference Vereecken, Dupuy and Rasmussen 23 – Reference Keski-Rahkonen, Kaprio and Rissanen 25 ) is more likely to have irregular breakfast consumption. However, cross-country comparative data on the prevalence of breakfast consumption or skipping, as well as on the association of SES with breakfast habits in children, is limited in Europe( Reference Vereecken, Dupuy and Rasmussen 23 ). In this context, the first results from the ENERGY project showed that children’s mean weekly breakfast consumption frequency in Europe was 5·9 d/week, ranging from 5·1 in Slovenia to 6·7 in Spain( Reference Van Lippevelde, Te Velde and Verloigne 26 ), with no significant differences between boys and girls( Reference Brug, van Stralen and Te Velde 8 ). Regarding ethnic background (assessed either by language spoken at home or by country of birth of the parents), the findings from the ENERGY project showed a higher prevalence of breakfast skipping in non-native compared with native children( Reference Brug, van Stralen and Chinapaw 7 ). These differences in breakfast habits among native and non-native children are possibly due to cultural differences regarding different dietary habits observed in families of different ethnic background( Reference van der Horst, Oenema and te Velde 27 ). Moreover, the fact that ethnic minority groups have higher rates of unhealthy behaviours might be due to their lower educational level( Reference Brug, van Stralen and Chinapaw 7 , Reference van der Horst, Oenema and te Velde 27 ). These differences may also be attributed to non-native children’s social and physical environment, since living in deprived environments or losing social context may prohibit children’s breakfast consumption( Reference Brug, van Stralen and Chinapaw 7 , Reference Magnusson, Hulthen and Kjellgren 28 ). None the less, the question of whether children’s breakfast consumption mediates the association between family sociodemographic characteristics and children’s weight status has not yet been investigated. Such an investigation can help to explain socio-cultural differences in children’s weight status and may provide entry points for interventions aiming to reduce socio-cultural inequalities in overweight and obesity.
Considering all of the above, the current study aimed to: (i) investigate the associations between certain family sociodemographic and parental characteristics (i.e. children’s gender, family structure, parental nationality, occupational status, education) with children’s breakfast consumption habits and overweight/obesity; (ii) examine the associations between children’s breakfast consumption habits and overweight/obesity; and (iii) assess the potential mediating effect of children’s daily breakfast consumption in the associations of family sociodemographic and parental characteristics with children’s overweight/obesity. These three aims will be jointly tested using mediation analyses.
Methods
Study design and participants
The rationale and organization of the ENERGY project( Reference Brug, te Velde and Chinapaw 29 ) and a comprehensive description of the design, procedures and methodology of the ENERGY school-based survey( Reference van Stralen, te Velde and Singh 30 ) are published elsewhere. Seven countries from the ENERGY Consortium, namely Belgium, Greece, Hungary, the Netherlands, Norway, Slovenia and Spain, participated in the cross-sectional survey. Switzerland joined the Consortium in a later phase( Reference Herzig, Dossegger and Mader 31 ). The school-based survey of the ENERGY project was carried out among 10- to 12-year-old children and their parents. The recruitment and data collection took place from March to July 2010 (Belgium, Greece, Hungary, the Netherlands, Norway, Slovenia and Spain) and between June and December 2010 (Switzerland). These countries were selected since they provide variation across regions in Europe and thus variation in potential obesogenic behaviours and prevalence of overweight and obesity. All participating countries obtained ethical clearance from the relevant ethical committees and ministries. The project adhered to the Helsinki Declaration and the Convention on Human Rights and Biomedicine of the Council of Europe.
Sampling was national in Greece, Hungary, the Netherlands and Slovenia. In Spain, schools in the region of Aragón were selected; Belgium selected schools from Flanders, Norway selected schools from the southern regions of the country, and Switzerland from the German-speaking part of the country( Reference van Stralen, te Velde and Singh 30 ). The sampling of schools was random, multistage and stratified by degree of urbanization in the regions under study. More details on the sampling procedure are presented elsewhere( Reference van Stralen, te Velde and Singh 30 ). A school recruitment letter was sent to the headmaster or principal of the participating schools, followed by a personal telephone call. Following the school’s approval for participation in the study, parents received a letter explaining the study purpose and were asked for written consent for their child’s and own participation. Detailed information on response rates at school, child and parent level has been reported elsewhere( Reference Brug, van Stralen and Te Velde 8 ). A total of 199 schools participated, with 7915 children (children’s response rate 60 %) and 6512 parents (parents’ response rate 55 %) completing questionnaires across the eight countries.
Data collection
Measurements in all countries were conducted according to a standardized protocol( Reference van Stralen, te Velde and Singh 30 ). The study entailed anthropometric measurements, a child questionnaire and a parent questionnaire. These questionnaires were used to measure children’s energy balance-related behaviours and potential individual, sociodemographic and environmental correlates of these behaviours. The data collection protocol and survey questionnaires for the ENERGY cross-sectional survey are available online (http://projectenergy.eu). Detailed information regarding the development, validity and reliability of the child questionnaire is published elsewhere( Reference van Stralen, te Velde and Singh 30 , Reference Singh, Vik and Chinapaw 32 ). Children completed the questionnaires addressed to them during school time. Children also received the parent questionnaire in a closed envelope to take home for completion by one of the parents.
Anthropometric measurements
Body height and weight measurements were conducted by trained research assistants. Children were weighed in light clothing without shoes using a Seca digital scale (Seca Alpha, model 861, Hamburg, Germany) with an accuracy of 0·1 kg. Body height was measured to the nearest 0·1 cm using a commercial stadiometer (Leicester Height Measure, Invicta Plastics Ltd, Oadby, UK) with children keeping their shoulders in a relaxed position, their arms hanging freely and their head aligned in the Frankfort horizontal plane. Two readings of each measurement were obtained. If the two readings differed more than 1 % then a third measurement was taken. BMI was calculated and the International Obesity Task Force (IOTF) cut-off points( Reference Cole, Flegal and Nicholls 33 , Reference Cole, Bellizzi and Flegal 34 ) were used to categorize participants into a dichotomous weight status variable, i.e. normal weight (also including underweight) and overweight/obese.
Breakfast consumption
Children’s breakfast consumption was assessed by two questions asking children on how many schooldays and how many weekend days per week they normally had breakfast. Breakfast frequency per week was calculated by adding up the answers of the two questions. Then, children were divided in two categories: (i) those with daily breakfast consumption (DBC); and (ii) those skipping breakfast at least once during weekdays (i.e. at least once out of five days) and/or weekends (i.e. at least once out of two days).
Family sociodemographic characteristics
Data on family sociodemographic characteristics obtained in the present study included parental educational status, employment status, family structure and nationality. The educational status was categorized as: ‘both parents <14 years of education’ and ‘at least one parent ≥14 years of education’, distinguishing families with at least one caregiver who had completed medium or higher vocational, college or university training from other families. The parental employment status was categorized as: ‘at least one unemployed’ and ‘both employed’. Family structure was divided into two categories: i.e. single- and dual-parent family. Regarding parental nationality, parents were further categorized as native and non-native based on their country of birth. A dichotomous variable was created, according to the definition of foreign ethnic background used by Statistics Netherlands, distinguishing children for whom both parents were born in the country of administration (native) from children for whom at least one parent was born in another country (non-native).
Statistical analyses
Categorical variables are summarized as relative frequencies (%), while continuous variables are presented as mean and standard deviation. Unadjusted associations between categorical variables were assessed using the χ 2 test.
Multilevel logistic regression analyses were performed for the overall sample, with children nested within classes, nested within schools (three-level random intercept model) in order to assess the statistical significance of the associations between family sociodemographic characteristics, children’s breakfast consumption and children’s overweight/obesity. More specifically, three multiple logistic regression models were applied to examine the simultaneous associations of all demographic characteristics (independent variables) with the probability of children’s DBC and childhood overweight/obesity (dependent variables). The results are presented as odds ratios and 95 % confidence intervals.
In addition, mediation of the associations of family sociodemographic characteristics with children’s weight status by children’s breakfast consumption was tested using a path analytic approach as outlined by MacKinnon et al.( Reference MacKinnon, Fairchild and Fritz 35 , Reference MacKinnon 36 ), although the cross-sectional design of the present study is not optimal for this kind of analysis. First, associations between each demographic characteristic and children’s overweight/obesity were examined (total association, path c). Second, associations of each demographic characteristic with children’s breakfast consumption (potential mediator) were studied (Action Theory test, path a). Third, the association between breakfast consumption of the children (mediator) and children’s overweight/obesity was examined (Conceptual Theory Test, path b) adjusted for the demographic variables. Finally, the associations between each demographic variable with children’s overweight/obesity adjusted for children’s breakfast consumption were estimated (direct path, path c′; Fig. 1).
Schematic presentation of the mediation models examined in the present study (DBC, daily breakfast consumption)
All coefficients (i.e. a, b, c and c′) obtained from the corresponding three-level random-intercept logistic regression model were standardized as recommended by MacKinnon and Dwyer because the outcome variable (overweight/obesity) and the mediator (children’s breakfast consumption) are dichotomous( Reference Mackinnon and Dwyer 37 ). The mediated effect (a×b) was calculated as the product of standardized coefficients a and b (indirect effect)( Reference MacKinnon, Fairchild and Fritz 35 ). This was calculated only when both path a and path b were significant. The statistical significance and the 95 % confidence interval of the mediated effects were estimated using Preacher and Hayes’ bootstrapping method( Reference Preacher and Hayes 38 ). In addition, the proportion mediated was calculated by dividing the indirect effect by the sum of the direct effect (standardized c′ coefficient) and the indirect effect (a×b). This was calculated only in the case of a significant total association (path c) to prevent unreliable and uninterruptable proportions.
All reported P values were based on two-sided tests. The level of statistical significance was set at P<0·05. The statistical software package STATA 12·0 was used for all statistical analyses. In particular, the command xtmelogit was used to apply the three-level random-intercept models.
Results
Descriptive data
In total, full data on the variables under investigation were available for 5444 children and their parents who filled in the relevant questionnaires in the eight participating countries. The population under study consisted of children and their parents having full data in all variables needed to test the research hypothesis (i.e. full anthropometric data and full data on children’s breakfast habits and family sociodemographic characteristics collected from questionnaires completed by children and their parents, respectively). In almost all countries participating in the ENERGY project full data were collected from the majority of children and their parents. More specifically, the response rates, taking into account these rates at the school, child and parent levels, ranged from 33·4 % in the Netherlands to 83·0 % in Spain. With the exception of the Netherlands where the low willingness to participate at the school level led to a relatively low response rate( Reference Brug, van Stralen and Te Velde 8 ), the relevant rates were not found to differentiate significantly among the other seven countries. Descriptive data on sociodemographic variables are presented for the total sample and by country in Table 1.
Family sociodemographic and parental characteristics for a cohort of 10- to 12-year-old children in Europe presented by country; the ENERGY (EuropeaN Energy balance Research to prevent excessive weight Gain among Youth) project, 2010
Descriptive data on parental nationality and parental education can be found elsewhere( Reference Brug, van Stralen and Chinapaw 7 , Reference Brug, van Stralen and Te Velde 8 ).
Family sociodemographic characteristics and children’s breakfast consumption
As presented in Table 2, overall 74 % of school-aged children reported DBC on weekdays and 87 % on weekends. DBC on weekdays varied from 56 % in Slovenia to 92 % in Spain and on weekends from 79 % in Greece to 93 % in Norway. The rates of DBC were higher on weekends than on weekdays in all participating countries. The highest DBC rates were observed in the Netherlands, Norway and Spain.
Percentage of 10- to 12-year-old children in Europe with daily breakfast consumption on weekdays and on weekends presented by family sociodemographic and parental characteristics and country; the ENERGY (EuropeaN Energy balance Research to prevent excessive weight Gain among Youth) project, 2010
a , bPercentages with the same superscript letter differentiate statistically significantly within each column (more specifically within each category of sociodemographic and parental characteristics presented in the table; P<0·05 derived from χ 2 test).
Moreover, in the total sample, the present study showed that DBC was significantly higher among children whose parents were native, were both employed and had higher educational status. As indicated in Table 2, depending on the country, respective significant differences in the prevalence of DBC on weekdays and/or on weekends between children with different family sociodemographic characteristics were almost consistently observed in Greece, Slovenia and Spain, while no significant differences were observed for Belgium, the Netherlands and Norway.
Associations of family sociodemographic characteristics with children’s breakfast consumption habits and children’s weight status
Multiple logistic regression analyses presented in Table 3 showed that boys (OR=1·25, 95 % CI 1·10, 1·43), children of native background (OR=1·53, 95 % CI 1·30, 1·80), with both parents employed (OR=1·20, 95 % CI 1·00, 1·43) and with at least one parent having more than 14 years of education (OR=1·48, 95 % CI 1·29, 1·69) were more likely to eat breakfast on weekdays, compared with their counterparts. On weekends boys were less likely than girls to have breakfast (OR=0·70, 95 % CI 0·60, 0·83), but children of native background (OR=1·40, 95 % CI 1·15, 1·71), with both of their parents employed (OR=1·58, 95 % CI 1·31, 1·89) and with at least one parent having more than 14 years of education (OR=1·33, 95 % CI 1·12, 1·58) were more likely to eat breakfast, compared with their counterparts. Boys were more like to be overweight/obese (OR=1·38, 95 % CI 1·20, 1·58) than girls, but children with both their parents employed (OR=0·81, 95 % CI 0·69, 0·96) and children with at least one parent having more than 14 years of education (OR=0·82, 95 % CI 0·71, 0·96) were less likely to be overweight/obese, compared with their counterparts.
Association between selected demographic characteristics and children’s breakfast consumption on weekdays and weekends and between selected demographic characteristics and overweight/obesity in 10- to 12-year-old children in Europe; the ENERGY (EuropeaN Energy balance Research to prevent excessive weight Gain among Youth) project, 2010. Results from multilevel multiple logistic regression*
DBC, daily breakfast consumption.
*Each one of the three multiple logistic regression analyses included and was thus adjusted for all independent variables presented in the table. Significant associations are indicated in bold font.
Prevalence of overweight/obesity by breakfast consumption habits
The prevalence of overweight/obesity was significantly higher among children skipping breakfast compared with those with DBC on weekdays in four countries (i.e. Belgium, Greece, Norway and Switzerland; Fig. 2) and on weekends in two countries (i.e. Norway and Spain; Fig. 3).
Prevalence of overweight/obesity in 10- to 12-year-old children in Europe, according to their breakfast consumption habits on weekdays (
, daily breakfast consumption (DBC) on weekdays; 
, skips breakfast on weekdays); the ENERGY (EuropeaN Energy balance Research to prevent excessive weight Gain among Youth) project, 2010. *P<0·05 for the comparison between DBC and breakfast skipping categories
Prevalence of overweight/obesity in 10- to 12-year-old children in Europe, according to their breakfast consumption habits on weekend days (
, daily breakfast consumption (DBC) on weekends; 
, skips breakfast on weekends); the ENERGY (EuropeaN Energy balance Research to prevent excessive weight Gain among Youth) project, 2010. *P<0·05 for the comparison between DBC and breakfast skipping categories
Mediation analyses
Table 4 summarizes the results derived from the mediation analyses. Regarding indirect associations (path c), gender, parental occupational and educational statuses were significantly associated with children’s overweight/obesity. Regarding path a, children’s DBC was found to be significantly associated with parental nationality, occupational and educational statuses. Furthermore, as concerns path b, children’s DBC was found to be significantly associated with children’s overweight/obesity after controlling for gender, parental nationality and educational status, respectively.
Association between selected demographic characteristics and children’s overweight/obesity and the potential mediating effect of children’s daily breakfast consumption (DBC) on weekdays and weekends on these associations in 10- to 12-year-old children in Europe; the ENERGY (EuropeaN Energy balance Research to prevent excessive weight Gain among Youth) project, 2010
N/A, not applicable. Figures in bold indicate statistically significant findings.
Coefficient c is the estimate of the association between demographic characteristic (X) and overweight/obesity (Y), overall or total association, adjusted for country and school.
Coefficient c′ is the estimate of the association between demographic characteristic (X) and overweight/obesity (Y), adjusted for the mediator (M); direct association.
Coefficient a is the estimate of the association between demographic characteristic (X) and children’s DBC (M).
Coefficient b is the estimate of the association of children’s breakfast consumption (M) and children’s overweight/obesity (Y) after controlling for independent variable.
Comp(a)=a×SD(X)/SD(M′).
Comp(b)=b×SD(M)/SD(Y″).
Comp(c)=c×SD(X)/SD(Y′).
Comp(c′)=c′×SD(X)/SD(Y″).
Var(Y′)=c
2×Var(X)+
$\pi$
2/3.
Var(M′)=a
2×Var(X)+
$\pi$
2/3.
Var(Y″)=c′2×Var(X)+b
2×Var(M)+2×b×c′×Cov(X,M)+
$\pi$
2/3.
Comp(a)×Comp(b) is the product-of-coefficient estimate; mediated effect.
95 % CI of Comp(a)×Comp(b) is 95 % confidence interval of the mediated effect.
$\pi$
2/3 is the variance of the standard logistic distribution.
Children’s DBC significantly mediated the associations between parental nationality and parental educational status with children’s overweight/obesity (path a×b). However, as the total association (path c/c′) between parental nationality and children’s overweight/obesity was not statistically significant, the proportion mediated by children’s DBC is reported only for the association between parental educational status and children’s overweight/obesity for which path c/c′ was statistically significant; this was found to be 17 % (Table 4).
Discussion
The objective of the current study was to investigate the associations of family sociodemographic characteristics with children’s weight status and whether these potential associations are mediated by children’s breakfast habits, using mediation analyses to jointly examine all objectives with a single type of analysis. To our knowledge, the present study is the first one examining the possible mediating role of one very important energy balance-related behaviour, i.e. breakfast consumption( Reference Douthwaite, Summerbell and Moore 39 ), in the association between family sociodemographic characteristics and children’s overweight/obesity. Further to that, the current study represents a research attempt to look deeper into the associations and interactions between social, behavioural and clinical variables pointed to by the theoretical framework upon which the conceptualization and design of the ENERGY project were based( Reference Brug, te Velde and Chinapaw 29 ). According to the results derived from the mediation analyses, children’s DBC was significantly related to all family sociodemographic characteristics examined (path a), with the exception of family structure. In particular, the current study showed that employment of both parents and higher educational status of at least one parent were positively associated with children’s DBC on weekdays and/or weekends. This finding confirms the associations between family SES indices and breakfast habits reported in earlier US and European studies( Reference Timlin, Pereira and Story 17 , Reference Vereecken, Dupuy and Rasmussen 23 – Reference Keski-Rahkonen, Kaprio and Rissanen 25 , Reference Vik, Bjornara and Overby 40 ). Furthermore, consistent with the findings of the Health Behaviour in School-aged Children (HBSC) study( Reference Vereecken, Dupuy and Rasmussen 23 ), the present study indicates that the association of family SES with children’s DBC shows regional differences that could reflect differences in socio-cultural norms regarding breakfast consumption.
The findings of the present study additionally show that children’s DBC varies widely within different European countries and between weekdays and weekends. More specifically, in only three out of eight countries do more than 80 % of children report eating breakfast daily on weekdays. However, children tend to consume breakfast more regularly on weekends, since in almost all countries more than 80 % of children report higher DBC on both weekend days. The results of the present study for weekdays are consistent with the results from the HBSC study indicating that daily consumption of breakfast varied from 33 % to 75 % in several European countries( Reference Vereecken, Dupuy and Rasmussen 23 ), as well as with those reported in a previous study using 24 h recall data from the ENERGY project( Reference Vik, Bjornara and Overby 40 ). Variations in breakfast consumption across countries may be explained by differences in cultural and socio-economic factors as well as in the availability of foods and breakfast served at school.
Regarding associations of family sociodemographic characteristics with children’s weight status, the current study reports lower likelihood of overweight and obesity in children whose parents were employed and had a higher level of education, as was also indicated by earlier papers that present results from the ENERGY project( Reference Brug, van Stralen and Te Velde 8 , Reference Vik, Bjornara and Overby 40 ). This finding is also in line with a recent review of previous cross-sectional studies( Reference Shrewsbury and Wardle 9 ) and could be partially ascribed to the less favourable dietary patterns (e.g. the higher consumption of energy-dense foods), the lower engagement in physical activity and the more time spent on sedentary activities by children of less educated parents( Reference James, Fowler-Brown and Raghunathan 41 – Reference Mazur, Marquis and Jensen 43 ). Additionally, children of higher-SES families tend to be more adherent and responsive to health-related recommendations and media messages compared with children of lower-SES families( Reference Adler, Boyce and Chesney 44 ).
Our finding concerning the negative association observed between children’s DBC and overweight/obesity is in line with those reported by earlier papers from the ENERGY project( Reference Van Lippevelde, Te Velde and Verloigne 26 , Reference Vik, Bjornara and Overby 40 ), as well as from longitudinal studies( Reference Timlin, Pereira and Story 17 , Reference Quick, Wall and Larson 45 ). An interpretation of these associations could be that DBC increases satiety and prevents consumption of high-energy-dense snacks later in the morning, thus favouring energy balance, which in turn helps maintenance of normal body weight( Reference Scaglioni, Agostoni and Notaris 46 ). Another possible explanation of our results could be that breakfast consumption can also be a proxy of a more general healthy lifestyle( Reference Sjoberg, Hallberg and Hoglund 47 ).
However, none of the previous studies reporting significant associations between family sociodemographic characteristics and children’s overweight/obesity have explored the potential mediating effect of breakfast consumption habits. The findings of the present study show that children’s DBC mediated the associations of parental nationality and educational level with children’s overweight and obesity. However, DBC was not a significant mediator of the associations between gender and children’s overweight/obesity, most likely due to the non-significant association between gender and DBC (path a). Furthermore, no significant mediating effect of DBC was observed in the association between parental occupational status and children’s overweight/obesity, and this is probably due to the borderline (P=0·053), but yet non-significant association between DBC and children’s overweight/obesity (path b). Further to the above, according to MacKinnon it is possible to have statistically significant a and/or b paths in the absence of a significant c path( Reference MacKinnon 36 ). This is confirmed by the findings of the present study, since in the case of the mediating effect of DBC in the association between parental nationality and children’s overweight/obesity path c was not statistically significant. This specific result showing a mediating effect of children’s DBC but no association between parental nationality and children’s weight status (the primary association of interest) is probably indicative of other possible factors (suppressors) influencing children’s body weight and/or of higher statistical power to detect a mediating than a total effect. In the case of the mediating effect of DBC in the association between parental educational level and children’s overweight/obesity, both c and c′ paths, as well as a and b paths, were statistically significant, which might indicate that other factors also play a role in explaining the association between parental education and children’s overweight status.
The major strengths of the present study are the large sample size obtained from different European countries, the use of a standardized protocol for data collection and data processing, and the objectively measured weight and height in children. However, there are also certain limitations. First and foremost, a causal relationship cannot be identified due to the cross-sectional design of the study. For instance, it may be that overweight children start skipping breakfast as a strategy to lose weight. Second, we used a rather crude dichotomous variable for DBC, i.e. distinguishing between respondents who reported to eat breakfast every day and those who reported to skip breakfast at least once weekly. This means that no distinction between (for example) children who skipped breakfast at least one day at weekdays and one day at weekends and those who never eat breakfast was made. The fact that, despite this crude measure, strong associations and mediation are found may suggest that these associations and mediation are likely to be even stronger if more detailed distinctions in breakfast habits could have been analysed. Third, the use of mediation analyses to test for possible mediating effect on associations between cross-sectional data, instead of longitudinal data, represents another limitation of the current study. Lastly, dietary behaviours are based on self-reports and may therefore be subjected to misreporting bias.
Conclusion
In conclusion, the present study showed that DBC partly mediated the association of parental education and ethnicity with overweight/obesity in their children. From a public health perspective, our findings suggest that promotion of DBC in families with non-native and less educated parents could be considered a promising component of childhood obesity preventive initiatives. Nevertheless, further studies – preferably intervention studies – are needed to provide additional evidence on these associations and on the mediating effect of children’s breakfast consumption habits.
Acknowledgements
Financial support: The current work has been carried out within the framework of the ENERGY project. The ENERGY project is funded by the Seventh Framework Programme (CORDIS FP7) of the European Commission, HEALTH (FP7-HEALTH-2007-B). The content of this article reflects only the authors’ views and the European Community is not liable for any use that may be made of the information contained herein. Conflict of interest: None. Authorship: Y.Μ. drafted the manuscript. Y.M., J.B., S.J.t.V., M.J.M.C., E.B. and I.D.B. contributed to the conception and design of the cross-sectional study. All authors were involved in revising and improving the drafts before submission. All authors approved the submitted manuscript. Ethics of human subject participation: All participating countries obtained ethical clearance from the relevant ethical committees and ministries. The project adhered to the Helsinki Declaration and the Convention on Human Rights and Biomedicine of the Council of Europe.
